Content
68%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A tight, example-first CLI reference that is genuinely executable for the common cases and well-sectioned. Its main weaknesses are the absent validation/verification pattern for destructive operations (delete, kill), an install path pointing outside the skill, and undocumented flags used in examples.
Suggestions
Add an explicit destructive-operation safety workflow, e.g. always run 'asset delete <ID> --dry-run --output json' first, confirm the preview, then execute and re-check with 'asset get <ID>'; do the same for 'session kill'. This also lifts the workflow_clarity cap of 3.
Fix or remove the 'cd agent-harness && pip install -e .' instruction — that directory is not part of the skill bundle — and document or drop the undocumented example flags ('--status failed', '--action delete', '--risk 5').
Consider moving the per-command-group detail and the four workflow recipes into one-level-deep reference files (e.g. workflows.md, commands/asset.md) to slim the single-file body of this broad CLI.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is command-driven and table-based with almost no concept explanation ('All commands support `--output json`' followed immediately by an example), but it has minor trimmable padding: the intro sentence restates the frontmatter description, 'Monitor active sessions / session list --active --output json' is repeated verbatim in workflows 3 and 4, and the `cli-anything-jumpserver ... --output json` prefix is spelled out many times. This sits above anchor 3 (mostly efficient) but short of anchor 5 (every token earns its place). | 4 / 5 |
Actionability | Nearly all guidance is copy-paste-ready command lines with concrete flags ('asset list --type host --output json', 'user create --name "New User" --username newuser --email user@example.com --output json --dry-run'), matching anchor 5's executable bar. It falls to 4 due to minor gaps: the install step 'cd agent-harness && pip install -e .' points at a directory that is not part of the skill bundle, and example flags like 'audit login --status failed', 'audit operate --action delete', and 'session command list --risk 5' are used without being documented anywhere in the body, so an agent cannot confirm they are valid. | 4 / 5 |
Workflow Clarity | The four 'Typical Agent Workflows' present a clear sequence (e.g., check connection -> list -> get details; create with --dry-run -> verify via 'perm users <PERM_ID>'), and the body states 'All mutation commands support `--dry-run` to preview without execution'. However, the CLI covers destructive operations (asset/user delete, session kill) and no workflow shows an explicit validation checkpoint or error-recovery loop for them — deletes and kills appear only as bare subcommands with no verify step — so per the rubric cap for destructive/batch operations without validation, this cannot exceed anchor 3. | 3 / 5 |
Progressive Disclosure | Sections are well-organized and appropriately scoped for an overview ('## Quick Start', '## Agent Usage Guidance', '## Output Formats', '## Exit Codes', '## File Locations'), with no nested or dead references, matching anchor 4's 'good structure; most content is appropriately placed'. It misses anchor 5 because no bundle files exist at all while the body still references an unbundled 'agent-harness' directory, and the per-group command detail plus the four workflow recipes are candidates for one-level-deep reference files given the CLI's breadth. | 4 / 5 |
Total | 15 / 20 Passed |